Construction and Empirical Analysis of Corporate Financial Crisis Early Warning Model under the Perspective of Multimodal Data Fusion
DOI:
https://doi.org/10.54097/f485xn92Keywords:
Financial crisis early warning, Multimodal data fusion, Deep learning, Attention mechanismAbstract
Corporate financial risks in the current economic environment are becoming more and more complex, and it is difficult to meet the early warning needs with single modal data analysis. In this study, we use deep learning technology to design a multimodal data fusion architecture and construct a corporate financial crisis early warning model that includes financial data, text information and market transaction data. The study innovatively introduces a two-way attention mechanism to achieve adaptive fusion of features, develops a multi-level interpretable analysis framework, and improves the accuracy of early warning while ensuring the transparency of the model, providing new ideas for enterprise risk prevention and control.
Downloads
References
[1] Zengyuan W U, Lingmin J, Xiangli H, et al. Research on Financial Crisis Early Warning Model for Foreign Trade Listed Companies Based on SMOTE-XGBoost Algorithm [J]. Journal of Computer Engineering & Applications, 2024, 60(11).
[2] Zhu L. Research on enterprise financial crisis early warning management based on PLS-BP [J]. Int. J. Wirel. Mob. Comput. 2023, 24:195-202.
[3] Zhang Z, Liu X, Niu H. Financial crisis early warning of Chinese listed companies based on MD&A text-linguistic feature indicators [J]. PLoS ONE (v.1;2006), 2023, 18(9):23.
[4] Hu Z, Wang L, Ding X, et al. Multimodal Data Fusion Defense Strategy for Campus Network Security: Research on Kolmogorov Arnold Networks Combined With B‐Spline Function [J]. Security & Privacy, 2025, 8(3).
[5] Li Z, Wang Q, Zhao Z. Research on small target detection in multispectral remote sensing images based on multimodal deep learning [J]. IOP Publishing Ltd, 2024.
[6] Zhang F, Cui M, Zhang C, et al. Research on precise identification and localization methods for static small targets based on multimodal data fusion [J]. Measurement, 2025, 239(000):115336.
[7] Systems M I. Retracted: The Construction and Empirical Research of College English Multimodal Teaching from the Perspective of New Media [J]. Mobile Information Systems, 2023, 2023(000):1.
[8] Zhao J, Zhang F, Gao L, et al. Revealing Daily Mobility Pattern Disparities of Monomodal and Multimodal Travelers through a Multi-Layer Cluster Analysis: Insights from a Combined Big Dataset [J]. Sustainability (2071-1050), 2024, 16(9).
[9] Shi W. Analysis of the value of news and cultural communication using multimodal learning [J]. Applied Mathematics and Nonlinear Sciences, 2024, 9(1).
[10] Shiri F, Guo X Y, Far M G, et al. An Empirical Analysis on Spatial Reasoning Capabilities of Large Multimodal Models [J]. 2024.
[11] Singh P, Kushwaha A K S, Varshney N. IMF-MF: Interactive moment localization with adaptive multimodal fusion and self-attention [J]. Journal of Intelligent & Fuzzy Systems, 2025.
[12] Aafjes-Van Doorn K, Girard J M .From intuition to innovation: Empirical illustrations of multimodal measurement in psychotherapy research [J]. Psychotherapy Research, 2025, 35(2):171-173.
[13] Liu Y, Liu Y, Qi Z, et al. TCNAttention-Rag: Stock Prediction and Fraud Detection Framework Based on Financial Report Analysis [J]. 2025.
[14] Wang J, Ding W, Zhu X. Financial Analysis: Intelligent Financial Data Analysis System Based on LLM-RAG [J]. 2025.
[15] Yang Y, Zhang X, Liu S, Du W. Tokyo Stock Exchange prediction with a hybrid model of LightGBM and DNN [C]//Proceedings of the 2nd International Conference on Mathematical Statistics and Economic Analysis (MSEA 2023). Nanjing, China, 2023.
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
